Process automation

Introduction

Zowie Decision Engine addresses a critical challenge in business automation: how to handle complex processes while maintaining natural, engaging customer interactions across all touchpoints.

Think of Decision Engine as an intelligent digital representative that can:

• Engage in natural, conversational interactions with your customers
• Execute your specific business processes and procedures accurately
• Connect seamlessly with your business systems to access and update information
• Walk customers through multi-step processes without creating confusion or friction
• Understand and respond appropriately to unexpected customer inputs

Unlike traditional automation tools that force customers into rigid paths or simplistic menus, Decision Engine can handle sophisticated interactions like managing subscriptions, processing returns, modifying orders, and updating account information—all while maintaining a conversation that feels natural and responsive across chat, email, and voice channels.

Technical Definition

Zowie Decision Engine is a process automation framework built on a modular architecture that integrates generative AI with deterministic business logic. It executes configurable workflows through a series of parameterized building blocks while maintaining standardized data representation through a semantic layer. The system enables programmatic orchestration of complex business processes with API-driven integration capabilities while providing natural language interfaces for end users.

Core Technical Components

Process Execution Framework

The Process Execution Framework orchestrates business processes through a series of interconnected building blocks:

1. User Input Gatherer: Collects and validates user information using Data Models for structure and validation.
2. User Message Block: Generates user responses either through LLM-generated content (with contextual guidelines) or direct text delivery for compliance scenarios.
3. Conditional Block: Creates branching paths using JavaScript-based condition evaluation against collected data and system states.
4. Code Block (Script Block): Executes custom JavaScript logic and API calls within a sandboxed runtime environment.
5. Call Another Process Block: Embeds sub-processes with shared data access through process reference and context passing.
6. Finish Block: Terminates process execution with configurable session state management and transition handling.

Data Models

Data Models standardize data representation and validation throughout the system:

1. Data Model Types: Supports single values, list values, and compound values with multiple data types (Text, Number, Date, Boolean, Option).
2. Validation Mechanisms: Implements both RegEx-based validation for structured data and LLM-based validation for complex inputs.
3. Technical Implementation: Uses JSON schema with in-memory data store, persistence layer, and just-in-time validation during collection.

Key details

Execution Flow

1. Process Initialization
   • User intent mapping to process definition
   • Context initialization
   • Building block loading and validation
2. Block Execution Cycle
   • Sequential block execution
   • Context updates based on execution results
   • Conditional path evaluation and navigation
3. External System Integration
   • Synchronous API calls through Code Blocks
   • Authentication handling
   • Response processing and error management
4. Context Management
   • Data persistence throughout conversation lifecycle
   • Scope management between main processes and sub-processes
   • Context reset conditions and handling

Multi-Channel Implementation

Decision Engine works across different communication channels through a channel-agnostic architecture:

1. Chat Implementation
   • Synchronous message processing
   • Rich UI element support where available
   • Session persistence with timeout handling
2. Email Implementation
   • Asynchronous message handling
   • Plain text content generation
   • Thread management and correlation
3. Voice Implementation
   • Speech-to-text integration
   • Natural language understanding adaptations
   • Context retention between voice interactions

Integration Approaches

Decision Engine integrates with existing business systems through multiple methods:

1. Direct API Integration
   • RESTful API support
   • OAuth and API key authentication
   • Request/response mapping to Data Models
2. Custom Integration Code
   • JavaScript execution environment
   • Library support for common protocols
   • Error handling and retry logic